Evaluating the Impact of Sequence Combinations on Breast Tumor Segmentation in Multiparametric MRI

TL;DR

The paper addresses how different mpMRI sequence combinations affect deep learning–based breast tumor segmentation. It systematically evaluates input configurations using nnU-Net on the multicenter BMMR2 dataset to segment functional tumor volume (FTV) and whole tumor mask (WTM), including generating FTV predictions to guide WTM segmentation. The main findings show that DCE_sub is effective for FTV segmentation ($DSC$ around $0.69 \\pm 0.18$), and that adding the predicted FTV with $DWI$ and $ADC$ improves WTM performance to about $DSC = 0.60 \\pm 0.21$, while adding T2w provides no significant gain. These results inform mpMRI protocol choices for segmentation tasks and lay groundwork for future work on predicting neoadjuvant chemotherapy response in breast cancer.

Abstract

Multiparametric magnetic resonance imaging (mpMRI) is a key tool for assessing breast cancer progression. Although deep learning has been applied to automate tumor segmentation in breast MRI, the effect of sequence combinations in mpMRI remains under-investigated. This study explores the impact of different combinations of T2-weighted (T2w), dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted imaging (DWI) with apparent diffusion coefficient (ADC) map on breast tumor segmentation using nnU-Net. Evaluated on a multicenter mpMRI dataset, the nnU-Net model using DCE sequences achieved a Dice similarity coefficient (DSC) of 0.69 $\pm$ 0.18 for functional tumor volume (FTV) segmentation. For whole tumor mask (WTM) segmentation, adding the predicted FTV to DWI and ADC map improved the DSC from 0.57 $\pm$ 0.24 to 0.60 $\pm$ 0.21. Adding T2w did not yield significant improvement, which still requires further investigation under a more standardized imaging protocol. This study serves as a foundation for future work on predicting breast cancer treatment response using mpMRI.

Figures (1)

• Figure 1: FTV and WTM segmentation examples using ConvLSTM-UNet, U-Net and nnU-Net with different sequence combinations in axial, coronal and sagittal views (from top to bottom). The segmentation examples are outlined on the first post-contrast DCE sequence for FTV segmentation and DWI (b-value=800$s/mm^2$) for WTM segmentation. The network segmentation is represented in green and the ground truth in red.